Monday, 8 January 2018
Exhibit Hall 3 (ACC) (Austin, Texas)
Self-consistent and temporally homogeneous long-term data sets of precipitation and temperature over the entire Great Lakes and Midwest regions are needed to provide inputs to hydrologic models, assess historical trends in hydroclimatic variables, and downscale global and regional-scale climate models. To support these needs, the hybrid gridded meteorological dataset at 1/16 degree resolution has been assembled over the Great Lakes and Midwest regions from 1915-2016 at daily time steps. Preliminary hydrologic simulations using the Variable Infiltration Capacity (VIC) hydrology model in watershed scale, however, showed significant underestimation of annual streamflow, resulting from the in situ precipitation gauge undercatch that is a very significant issue throughout this region. Undercatch corrections of individual stations is generally infeasible in the U.S., due to missing metadata and very limited information regarding gauge type and measurement protocols (especially for snow). As an alternative to this approach, we developed macroscale post processing techniques to adjust our regridded precipitation product from 1950-2016 forwards, accounting for undercatch as a function of wind speed (WS) simulations obtained from NCAR reanalysis. For snow, comparison of the snow to rain ratio for a group of high-quality stations allowed us to use variable correction factors in different parts of the domain to account for differing snow measurement protocols. To adjust pre-1950 data, monthly correction factors based on post 1950 were applied to pre-1950 data. Forty watersheds listed in Hydro-Climate Data Network (HCDN) were used to compare simulated and observed streamflow to evaluate the gridded meterological datasets with and without undercatch corrections. Observed WS data at airport stations were also compared to NCAR reanalysis based WS products that we used for precipitation correction.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner